Archive for July, 2006

Think of something! A computer guesses what you are thinking of, usually in 20 questions or less, at 20q.net.

We call this kind of application “Artifical Intelligence“, and yet there’s not much intelligence at work here. The computer is simply accumulating and classifying previous responses, then using a neural network to tie them all together. And yet, the results are impressive!

The computer easily guessed fire, frisbee, grass, lover, car, freeway, tennis, etc. But I stumped it with “researcher”, a word not in its knowledgebase. Here’s how it went:

• Is it classified as Animal, Vegetable or Mineral? Animal
• Can you buy it at a store? No
• Does it eat leaves? Sometimes
• Is it very large? No
• Does it live in trees? No
• Is it a pure bred animal? No
• Does it make noise? Yes
• Can you control it? Usually
• Is it small? No
• Is it something you bring along? No
• Is it outside? Sometimes
• Is it comforting? Sometimes
• Is it considered intelligent? Yes
• Does it perform? Probably
• Does it play music? No
• Can it carry people? Rarely
• Does it have a long tail? No
• Can you hold it? Maybe
• I am guessing that it is a family? Wrong
• I am guessing that it is people? Wrong
• Is it involved in movies? No
• Can it be used for recreation? Sometimes
• Would you find it in an office? Sometimes
• I am guessing that it is a colleague? Wrong
• Would you like to be one? Yes
• Can you find it in a church? Sometimes
• Do you use it in public? Yes
• Can you use it with your friends? Yes
• I am guessing that it is a football team? Wrong
• I am guessing that it is a jester? Hmmm…

Some of my answers above were a bit dubious. You might get different results with different answers - and you can also play in different languages, with a distinction made between American English and British English.

You can also find this application as the “20Q” child’s toy.

(via tlspiegel-ga and pafalafa-ga)

I’ve posted here before about an absolutely wonderful resource for historical research, newspaperarchive.com.

Unfortunately, though, it’s a subscription service, which puts it out of reach of a lot of would-be users, so I’m glad to make note of some good news.  Starting immediately, newspaperarchive.com is making itself available to schools and libraries at no charge.

This is such a wonderful, useful, valuable service that you’d be wise to urge your school or local library to sign up, pronto.

 pafalafaga Dave Sarokin

The Portugese version of Wikipedia has a tiny stub of an article about Google Answers. Here’s a rough translation:

Google Answers is a service offered by Google which allows the users to submit any question and get an answer, by paying 2.50 dollars. An answer to a submitted question is edited by the community itself and sent to the user who asked it…

Clearly this needs some improvement, because the price is variable (from $2 to $200 plus the fifty cent listing fee), and because the paid answer is provided by a “carefully screened researcher” rather than “the community itself”.

So if you have the Portugese skills, please go ahead and edit the article. Anyone can edit Wikipedia!

(via politicalguru-ga and guillermo-ga)

Julian Beever is an artist with an incredible eye for perspective - so good that he can mentally invert the perspective, leading to some amazing pavement art.

First, he draws a scene with an intentionally-distorted perspective, like this:

Then, when you view it from the other direction, it appears to jump into three dimensions:

There are many more examples of Julian Beever’s amazing anamorphic illusions on his own website, and you can read more about the artist and his work in Wikipedia.

Which Google Answers Researcher has answered the most questions?

That would have to be GAR powerhouse bobbie7-ga, with 3192 questions answered.

Next in line seems to be the veritable pinkfreud-ga, with 2190 questions answered, followed closely by juggler-ga on 2045.

Other high rankers are tutuzdad-ga on 1704, pafalafa-ga on 1532, scriptor-ga on 1464 and  justaskscott-ga on 1107.

Have I missed anyone who has answered over a thousand questions?

I’ve answered a whopping 187.

I guess most Google Answers Researchers have used the Internet Archive’s Wayback Machine, but if you haven’t then it’s definitely one to check out.

Start at the Internet Archive Home Page, and type a URL into box above the “Take me back” button (not the “Search” box at the top). You will then see a page listing every date for which the Internet Archive holds a copy of that webpage. Just click on the date that you are interested in.

For example, here’s how the Google home page looked in 1998.

The advanced search lets you find archived pages in other ways, and also allows you to highlight the changes from one date to another.

How many Google Answers researchers are there? If you go to the Google Answers Home Page it says that “More than 500 carefully screened Researchers are ready to answer your question” (you don’t see that message when you are logged in).

How many of those researchers have answered questions? Knowledge_seeker-ga counted 57 who had answered one or more questions within the last 60 days.

Researchers come and go. I have answered quite a few questions recently, yet I’ve been very inactive at times. Twice I have received a December check from Google for a mere $1.50, for answering just one $2 question. It seems that on December 31st the accountants like to square the books.

Anyway, back in 2004 Benjamin Edelman of Harvard University studied Earnings and Ratings at Google Answers. He checked all the answers posted from the start of the service in April 2002 until November 2003, and found 534 distinct question answerers.

There is a lot of other fascinating information in Edelman’s study. For example, did you know that tipping correlates with number of links provided in the answer? Edelman also found that earnings increased with researcher experience - he estimated that every extra question a researcher answers raises their earnings by about two cents per hour.

Coming back to those 534 researchers: what else do we know about them? We know they come from many parts of the globe - aceresearcher-ga once assembled a partial list and found researchers from Argentina, Australia, Austria, Barbados, Belgium, Brazil, Canada, Chile, China, Colombia, Cyprus, Czech Republic, England, Germany, Hungary, India, Ireland, Israel, Italy, Japan, Lebanon, Malaysia, Netherlands, New Zealand, Norway, Philippines, Poland, Romania, Russia, Scotland, Singapore, South Africa, Spain, Switzerland, Tanzania, Turkey and the United States.

Because it’s a slow day here and has been raining since morning, I figured I’d do a bit of counting myself. I did a quick hand count of how many of the “600+” researchers have been actively answering questions for the past 60 days.

Answer: 57

Of those, 54 were active in the last 30 days.

Of those, during the last 30 days:

1 Answered 75 questions

1 Answered 50 questions

1 Answered 44 questions

10 Answered 20-30 questions

13 Answered 10-19 questions

The rest answered fewer than 10 questions apiece.

These numbers are very much in keeping with a similar count I did about 2 years ago. I no longer have the exact numbers, but I remember the total number of active researchers being just around 60, with a small subset of those picking up a high percentage of the workload.

I leave it to commenters to decide what this all means …

-K~

What happens to a question after it is posted to Google Answers? Here’s a rough idea.

Most of this is wild extrapoliation from my own experience; the rest is based on quick approximate counts from publicly-accessible pages at Google Answers.

• One person posts the question
• A few percent of questions are removed by the Google Answers Editors because they contravene the Terms of Service for Google Answers - for example, because they include identifiable personal information
• A few percent of questions are accidentally posted more than once. Researchers usually post a comment to alert other researchers to the duplicate question
• Over 600 researchers have been approved by Google to answer the question
• Around a hundred researchers will read the question title, of whom…
• Perhaps twenty researchers will read the question body.
• Half a dozen researchers will lock the question to conduct some preliminary research, of whom
• Two will research the question in depth, of whom
• One will ask for clarification, after which
• There will be an exchange of clarification details of between zero and a dozen or more messages (averaging around three).
• Finally, for about 50% of clarified questions,
• One researcher will post an answer.

This will result in:

• Between zero and a dozen or more request for clarification and clarifications, averaging two or three, plus
• Between zero and over a thousand comments, mostly from non-researchers.

The question will then:

• Be rated in about one-third of cases: usually five stars, sometimes four stars, occasionally three stars, rarely two stars and occasionally one star
• Be tipped in about 25% of cases, usually from $1 to $10 but can be up to $100
• Be rejected by the customer, or withdrawn by the researcher, in under 5% of cases

Of the unanswered questions:

• A small number will be closed early, but
• The vast majority will expire normally.

Of the answered questions:

• Under five percent will be referenced and linked to from a future question

Does anyone have any other metrics of interest to suggest?

Researcher cynthia-ga found this ASCII owl in her web wanderings:

{o,o}
|)__)
-”-”-

But the genre goes back further than that. Modernmechanix.com is featuring an article from a 1948 issue of Popular Mechanics that shows an owl created with typewriter art:

Here’s another, from http://www.asciipr0n.com/m1a9366b/ascii/owl.html (I didn’t hyperlink the URL because some of the other ASCII Art at that site is adult-oriented):